How to put values to the new separate DataFrame columns based on the value of specific already exist-CodePudding

I have a DataFrame with the mixed data of the different sources, note, that there a portions of data obtined at the same timestamp:

 -------------------------------------- ------ ------------------- ----------------- --------------- ----------------------- 
|devicename                            |value |time               |one_type_id|another_type_id|write_time             |
 -------------------------------------- ------ ------------------- ----------------- --------------- ----------------------- 
|Real_Power_KPI                        |0.0   |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:36.129|
|Voltage_Sensor                        |243.93|2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:36.129|
|Current_Sensor                        |0.0   |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:36.129|
|Casing_Vibration_Sensor               |0.0   |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:36.369|
|Water_Temperature_Sensor              |17.0  |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:36.369|
|Environment_Ambient_Temperature_Sensor|17.0  |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:36.369|
|Pump_Vibration_Sensor                 |0.0   |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:36.369|
|Water_Level_Sensor                    |15.0  |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:36.369|
|Environment_Humidity_Sensor           |81.2  |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:36.369|
|Water_Temperature_Sensor              |17.0  |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |
|Casing_Vibration_Sensor               |0.0   |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |
|Pump_Vibration_Sensor                 |0.0   |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |
|Environment_Ambient_Temperature_Sensor|17.0  |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |
|Water_Level_Sensor                    |15.0  |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |
|Environment_Humidity_Sensor           |81.2  |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |
|Real_Power_KPI                        |0.0   |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |
|Voltage_Sensor                        |245.01|2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |
|Current_Sensor                        |0.0   |2021-03-24 07:06:35|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |
|Real_Power_KPI                        |0.0   |2021-03-24 07:06:36|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |
|Voltage_Sensor                        |244.31|2021-03-24 07:06:36|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |
|Current_Sensor                        |0.0   |2021-03-24 07:06:36|NP20100000       |NP20100000     |2021-03-24 07:06:37.01 |

so, what i want is to have separate columns for Real_Power_KPI, Voltage_Sensor, Current_Sensor with their corresponding values joined in one row, while having the same timestamp.

Something like

|timestamp          |Real_Power_KPI|Voltage_Sensor|Current_Sensor|
|2021-03-24 07:06:36|0.0           |244.31        |0.0           |

so how I can do this transpose operation the most optimal way?

UPD.

In 过过招's answer the Python code is proposed, below is the Scala for that:

val df = dailySensorData.filter("devicename in ('Real_Power_KPI', 'Voltage_Sensor', 'Current_Sensor')") 
    .groupBy("time", "devicename").agg(expr("sum(value) as total")) 
    .groupBy("time").pivot("devicename").agg(expr("first(total)"))
df.show(false)

CodePudding user response：

Group and summarize first, and then use pivot to turn rows into columns.

df = df.filter("devicename in ('Real_Power_KPI', 'Voltage_Sensor', 'Current_Sensor')") \
    .groupBy('time', 'devicename').agg(F.expr('sum(value) as total')) \
    .groupBy('time').pivot('devicename').agg(F.expr('first(total)'))
df.show(truncate=False)